Title: AI-Generated Traces for Novice Programmers: Learning Effects and Learner Differences in a Multi-Institutional Study

Abstract

Introductory programming (CS1) courses frequently encounter difficulties in helping students comprehend how programs execute. Although visualizations can render execution processes explicit, their efficacy is heavily influenced by design and context. Furthermore, empirical data regarding the utility of AI-generated visualizations remains scarce. To address this, we introduce Generated Animated Traces (GATs)—AI-driven animations that blend source code, execution states, and conceptual analogies into a narrated format.

We executed a multi-institutional study within CS1 courses at two universities, comparing the impact of GATs against traditional textual explanations. The study involved 961 students in Python courses and 151 students in Java courses. Our assessment metrics included immediate learning outcomes, user experience, end-of-course engagement, and final exam performance.

The results indicate that while GATs offer selective advantages for immediate learning, these benefits are both context-dependent and transient. Crucially, we found that learner engagement profiles moderate the effect of GATs on performance. This insight highlights the critical need for personalized educational strategies.

Source: arXiv Generated at: 2026-06-03 00:00:00 UTC